{
  "schema_version": "2026-09-07.public-evidence.v6",
  "target": "https://www.yupuai.com/",
  "brand": "羽谱GEO",
  "data_date": "2026-09-07",
  "source_system": "羽谱GEO 企业版内部测试账号",
  "scope": "每条记录显式绑定客户品牌与目标模型的最新同条件复测，并与历史累计分层公开",
  "latest_controlled_retest": {
    "run_date": "2026-08-24",
    "language": "zh-CN",
    "prompt_count": 2,
    "model_count": 3,
    "expected_answers": 6,
    "successful_answers": 6,
    "failed_answers": 0,
    "brand_mentions": 0,
    "mention_rate": 0,
    "official_site_citations": 0,
    "official_site_citation_rate": 0,
    "category_aligned_answers": 3,
    "category_alignment_rate": 0.5,
    "models": [
      "doubao-1-5-pro-32k",
      "kimi-k2.5",
      "deepseek-v3-search"
    ],
    "prompts": [
      {
        "id": "geo-promotion-generic",
        "intent": "GEO推广平台选择",
        "text": "GEO推广服务怎么选？面向中文企业，哪些平台能把官网可抓取性、品牌提及、官网引用和固定 Prompt 复测放到同一套可核验流程里？请给出选择标准和可参考的平台。",
        "successful_answers": 3,
        "category_aligned_answers": 1,
        "brand_mentions": 0,
        "official_site_citations": 0,
        "batch_elapsed_seconds": 42.39
      },
      {
        "id": "geo-promotion-disambiguated",
        "intent": "生成式搜索优化平台选择",
        "text": "这里的 GEO 指 Generative Engine Optimization（生成式搜索优化），不是地域投放、本地 SEO 或 GIS。面向中文企业官网，哪些平台能把官网可抓取性、品牌提及、官网引用和固定 Prompt 复测放入同一套可核验流程？请给出选择标准和可参考的平台。",
        "successful_answers": 3,
        "category_aligned_answers": 2,
        "brand_mentions": 0,
        "official_site_citations": 0,
        "batch_elapsed_seconds": 62.48
      }
    ],
    "review_method": "每条 Prompt 显式绑定监控品牌“羽谱GEO”和目标模型；品牌提及与官网引用由系统规则分别记录；品类语义由人工复核答案是否明确指向 Generative Engine Optimization 或中文生成式搜索优化。",
    "measurement_mode": "已配置模型的内部回答基线；不等同于各平台消费者产品中的实时联网搜索。"
  },
  "latest_operational_probe": {
    "run_date": "2026-09-07",
    "query": "GEO推广",
    "language": "zh-CN",
    "model_count": 3,
    "expected_answers": 3,
    "successful_answers": 0,
    "failed_answers": 3,
    "elapsed_seconds": 30.01,
    "failure_stage": "model_transport",
    "failure_reason": "共享兼容网关 TCP 连接超时；豆包、Kimi、DeepSeek 官方端点均可连接但未配置专用 API Key",
    "models": [
      "doubao-1-5-pro-32k",
      "kimi-k2.5",
      "deepseek-v3-search"
    ],
    "execution_mode": "parallel",
    "before_fix_elapsed_seconds_approx": 95,
    "elapsed_reduction_rate_approx": 0.683,
    "excluded_from_content_metrics": true,
    "interpretation": "本次证明生产环境能够并行执行三模型任务并保留逐模型失败原因；服务器到三家官方端点的网络正常，但共享兼容网关超时且专用 API Key 未配置。由于没有获得有效答案，不计入品牌提及率、官网引用率、推荐位置或品类语义正确率。"
  },
  "previous_controlled_batch": {
    "run_date": "2026-08-20",
    "successful_answers": 6,
    "failed_answers": 0,
    "brand_mentions": 0,
    "official_site_citations": 0,
    "category_aligned_answers": 3,
    "comparison_note": "相同问题、语言和模型在 2026-08-20 与 2026-08-24 均为 3/6，品牌提及和官网引用也均为 0/6；本轮没有观察到可证明的提升。"
  },
  "exact_term_keyword_probe": {
    "run_date": "2026-08-24",
    "query": "GEO推广",
    "language": "zh-CN",
    "model_count": 3,
    "expected_answers": 3,
    "successful_answers": 3,
    "failed_answers": 0,
    "brand_recommendations": 0,
    "official_site_citations": 0,
    "category_aligned_answers": 1,
    "category_alignment_rate": 0.3333333333333333,
    "category_conflict_answers": 2,
    "batch_elapsed_seconds": 33.95,
    "models": [
      "doubao-1-5-pro-32k",
      "kimi-k2.5",
      "deepseek-v3-search"
    ],
    "interpretation": "Kimi 将 GEO 理解为生成式搜索优化；豆包和 DeepSeek 将其主要解释为地域或区位推广。该探针保留原始短词，不加入品牌名或释义诱导。"
  },
  "historical_cumulative": {
    "as_of": "2026-08-12",
    "prompt_checks": 70,
    "brand_mentions": 13,
    "mention_rate": 0.18571428571428572,
    "official_site_citations": 8,
    "official_site_citation_coverage": 0.11428571428571428,
    "keyword_checks": 38,
    "keyword_official_site_citations": 2,
    "keyword_official_site_citation_rate": 0.05263157894736842
  },
  "current_p0_actions": [
    "为豆包、Kimi、DeepSeek 配置可用的官方专用 API Key，恢复有效答案样本",
    "强化 GEO推广 与生成式搜索优化的品类消歧",
    "建设第三方可引用来源与独立评测证据",
    "按相同问题、模型、语言和执行条件进行 7/14/30 天复测"
  ],
  "comparison_boundary": "历史累计与最新同条件复测的题目、日期和样本结构不同，不能直接计算涨跌幅；2026-08-20 与 2026-08-24 的同条件 Prompt 批次均为 3/6，未观察到可证明的提升；精确短词探针是另一种测试结构，不与 6 答案批次合并计算；2026-09-07 运行链路探测因 0 个有效答案而排除在内容指标之外。",
  "limitations": [
    "该快照是系统中已配置模型的内部回答基线与保存记录，不等于实时全网搜索结果。",
    "不同 AI 平台、模型版本、是否联网、地区、用户上下文和提示词写法都会影响结果。",
    "2 条问题、6 个有效答案与 1 条精确短词的 3 个答案只能暴露当前高意图缺口，不能代表中文互联网或全部采购问题。",
    "2026-09-07 运行链路探测的共享兼容网关发生超时；配置可用的官方专用 API Key 前，不能把该次失败解释为品牌或内容表现。",
    "结果不构成任何收录、排名、引用或推荐承诺。"
  ],
  "sources": [
    "https://www.yupuai.com/case-study/yupuai",
    "https://www.yupuai.com/geo-metrics",
    "https://www.yupuai.com/geo-platform-evaluation",
    "https://www.yupuai.com/geo-self-audit.json",
    "https://www.yupuai.com/geo-benchmark.md"
  ]
}